sscdf

Binary sparse storage scheme for SuiteSparse::GraphBLAS utilizing netCDF4

The format specification is found here.

Single Object Storage

sscdf contains a convenience method to save a single object.

write(file, obj, comment=None)

• file can be a filename or a file-like object opened in "write" mode.
• obj is the graphblas object (Matrix, Vector, or Scalar)
• comment is optional

Single Object Loading

Loading the primary object can be done with a convenience method.

read(file)

• file can be a filename or a file-like object opened in "read" mode.

This returns the primary object in the file, reconstituted as a graphblas object (Matrix, Vector, or Scalar).

Multiple Object Storage

When writing multiple associated graphblas objects to a single sscdf file, the preferred method is using a Writer in a with block.

with sscdf.Writer(filename) as w:
    w.write(my_matrix, comment="Data came from the internet")
    w.write(my_matrix.reduce_rowwise(), name="row_degrees")

Multiple Object Loading

When a sscdf file contains multiple objects, the easiest way to access the named objects is using a Reader in a with block.

with sscdf.Reader(filename) as r:
    my_matrix = r.read()
    row_degrees = r.read('row_degrees')

Inspecting sscdf Files

To view what objects exist in an sscdf file, use the info() method. This will show the metadata and attributes of the primary as well as any secondary objects in the file.

>>> sscdf.info(filename)
{'version': '1.0',
 'format': 'CSR',
 'shape': [5, 5],
 'data_types': {
     'pointers_0': 'uint64',
     'indices_1': 'uint64',
     'values': 'float32'},
 'comment': 'Data came from the internet',
 'SECONDARY': {
     'row_degrees': {
         'version': '1.0',
         'format': 'VEC',
         'shape': [5],
         'data_types': {
             'indices_0': 'uint64',
             'values': 'float32'},
     }
 }
}